Identifying spoken commands by templates of ordered voiced and unvoiced sound intervals

Abstract

A method is disclosed for identifying a spoken command by detecting intervals of voiced and unvoiced sound, and then comparing the order of voiced and unvoiced sounds to a set of templates. Each template represents one of the predetermined acceptable commands of the application, and is associated with a predetermined action. When the order of voiced and unvoiced intervals in the spoken command matches the order in one of the templates, the associated action is thus selected. Silent intervals in the command may also be included for enhanced recognition. Efficient protocols are disclosed for discriminating voiced and unvoiced sounds, and for detecting the beginning and ending of each sound interval in the command, and for comparing the command sequence to the templates. In a sparse-command application, this method provides fast and robust recognition, and can be implemented with low-cost hardware and extremely minimal software.

Description

FIELD OF THE INVENTION[0001]The invention relates to voice-activation technology, and particularly to means for recognizing a spoken command by detecting time intervals containing voiced and unvoiced sound.BACKGROUND OF THE INVENTION[0002]Voice-activation technology is a rapidly evolving field. Fascinating applications appear almost daily. Prior art in this field is primarily directed toward the interpretation of free-form speech such as dictation and general questions. Most of the emerging applications, however, involve relatively simple devices that perform just a few specific operations. Desirable products that could be fully operated with a few predetermined commands include consumer devices (games, hobby devices, counters and timers, kitchen gadgets, home automation, exercise and sporting applications, toys, learning aids, products for the disabled), industrial systems (hands-free system interfaces, security monitoring, semi-autonomous machining and assembly, devices for rapid ...

AI Technical Summary

Benefits of technology

The invention is a method for recognizing spoken commands using an integrated signal that combines digital measurements. The method allows for the rapid and reliable identification of commands using a simple process that requires minimal software and hardware. The method can be used in a range of devices and applications, making it more cost-effective than previous methods. Overall, the invention enables a new level of efficiency and reliability for voice-activated applications.

Problems solved by technology

Unfortunately, the prior art serves these applications poorly.

Unfortunately, many commands that consumers expect to use cannot be discriminated on frequency alone, and many command words include brief phonemes for which frequency analysis is ambiguous at best.

Consumers are sensitive about their user interface, and do not gladly tolerate awkward commands or arbitrary limitations.

Converting the sound signal to a frequency spectrum is slower and more expensive than processing the sound as it arrives, in real-time.

For example, the dual-bandpass system of Ariav requires multiple gain stages with multiple filter components, and then the low- and high-frequency channels must be digitized separately, all of which increases board cost and software complexity.

Perhaps one could digitize an unfiltered signal instead, and then use Fourier analysis to separate the two frequency components; but this would require a greatly expanded processor and memory, negating any savings.

Moreover, key features of the sound wave are lost in a conventional FFT because it displaces phase information.

The frequency domain is a valid representation of sound only with complex-number or vector Fourier transformation, requiring even larger processors and memories, with costs that more than offset any other savings.

Images